It is proposed that preseismic fault creep may be the underlying process that is responsible for observations of earthquake precursors. The assertion that fault creep precedes earthquakes is supported by evidence from at least some earthquakes and by analogy with detailed laboratory observations. Laboratory observations of stick slip reveal that at least two stages of preseismic slip are an intrinsic part of the process leading to seismic slip on preexisting faults with inhomogeneous stress or strength. During the slowly propagating first stage of creep it is assumed that the length of the creeping fault segment is proportional to the source length of the subsequent earthquake. The data giving the well-known relationship between precursor time and earthquake magnitude are closely satisfied if the rate of propagation of the first stage of creep is independent of fault length. Long-term precursors may arise because of stress-strain variations during the first stage of fault creep. Observations of short-term precursors immediately prior to earthquakes may be related to the second short-lived state of preseismic fault slip seen in stick slip experiments.